Computing the Exact Pareto Front in Average-Cost Multi-Objective Markov Decision Processes — ThinkLLM